Fuzzy model based symptom generation and fault diagnosis for nonlinear processes

In this contribution, local linear fuzzy models are used for fault detection and fault diagnosis (FDD) for nonlinear processes. A Takagi-Sugeno type fuzzy model of the nominal process is identified off-line and linearized in the current operating point. In addition, a second linear model is identified on-line by applying a recursive least-squares (RLS) algorithm. The deviation in the parameters of both models lead to symptoms which indicate the state of the system. The approach enables FDD in all operating regimes. The approach is successfully applied to a electro-pneumatic valve with connected pipe system. Here, four symptoms were generated out of two measurements and six faults can be detected. In order to model the symptom fault causality, a MLP classification structure is implemented.